by john laufer national bureau of standards - nasa · by john laufer national bureau of...
TRANSCRIPT
REPORT 1174
THE STRUCTURE OF TURBULENCE IN FULLY
DEVELOPED PIPE FLOW
By JOHN LAUFER
National Bureau of Standards
!"
https://ntrs.nasa.gov/search.jsp?R=19930092199 2018-07-15T09:33:34+00:00Z
NOTICE
THIS DOCUMENT HAS BEEN REPRODUCED
FROM THE BEST COPY FURNISHED US BY
THE SPONSORING AGENCY. ALTHOUGH IT
IS RECOGNIZED THAT CERTAIN PORTIONS
ARE ILLEGIBLE, IT IS BEING RELEASED
IN THE INTEREST OF MAKING AVAILABLE
AS MUCH INFORMATION AS POSSIBLE.
F
!
o
f
National Advisory Committee for Aeronautics
Headquarters, 1512 H Street NW., Washington 25, D. C.
Created by act of Congress approved March 3, 1915, for the supervision and direction of the scientific study
of the problems of flight (U. S. Code, title 50, sec. 151). Its membership was increased from 12 to 15 by act
approved March 2, 1929, and to 17 by act approved May 25, 1948. The members are appointed by the President,and serve as such without compensation.
JEROME C. HUNSAKER_ SC. D., Massachusetts Institute of Technology, Chairman
DETLEV W. BRONK, PH.D., President, Rockefeller Institute for Medical Research, Vice Chairman
JOSEPH P. ADAMS, LL.D., member, Civil Aeronautics Board.
ALLEN V. ASTIN, Pm D., Director, National Bureau of Standards.
PRESTON R. BASSETT, M. A., President, Sperry Gyroscope Co.,Inc.
LEONARD CARMICHAEL, Pm D., Secretary, Smithsonian Insti-tution.
RALPH S. DAMON, D. Eng., President, Trans World Airlines, Inc.
JAMES H. DOOLITTLE, Sc. D., Vice President, Shell Oil Co.
LLOYD HARRISON, Rear Admiral, United States Navy, Deputyand Assistant Chief of the Bureau of Aeronautics.
RONALD M. HAZEN, B. S., Director of Engineering, Allison
Division, General Motors Corp.
RALPH A. OFSTIE, Vice Admiral, United States Navy, Deputy
Chief of Naval Operations (Air).
DONALD L. PUTT, Lieutenant General, United States Air Force,Deputy Chief of Staff (Development).
DONALD A. QUARLES, D. Eng., Assistant Secretary of Defense(Research and Development).
ARTHUR E. RAYMOND, SC. D., Vice President--Engineering,Douglas Aircraft Co., Inc.
FRANCIS W. REIeHELDERFER, SC. D., Chief, United StatesWeather Bureau.
OSWALD RYAN, LL.D., member, Civil Aeronautics Board.
NATHA_¢ F. TWINING, General, United States Air Force, Chiefof Staff.
HUGH L. DRYDEN, PH. D., Director
JOHN W. CROWLEY, JR., B. S., Associate Director for Research
JOHN F. VICTORY, LL.D., Executive Secretary
EDWARD H. CHAMBERLIN, Executive O_cer
HENRY J. E. REID, D. Eng., Director, Langley Aeronautical Laboratory, Langley Field, Va.
SMITH J. DEFRANCE, D. Eng., Director, Ames Aeronautical Laboratory, Moffett Field, Calif.
EDWARD R. SHARP, Sc. D., Director, Lewis Flight Propulsion Laboratory, Cleveland Airport, Cleveland, Ohio
LANGLEY AERONAUTICAL LABORATORY AMES AERONAUTICAL LABORATORY LEWIS FLIGHT PROPULSION LABORATORY
Langley Field, Va. Moffett Field, Calif. Cleveland Airport, Cleveland, Ohio
Conduct, under unified control, for all agencies, of scientific research on the fundamental problems of flight
.... .(
,i
• r-- •_- ....
II
J
REPORT 1174
THE STRUCTURE OF TURBULENCE IN FULLY DEVELOPED PIPE FLOW 1
By JOHN LAUFER
SUMMARY
_J/leasurements, principally with a hot-wire anemometer, weremade in fully developed turbulent flow in a lO-inch 19ipe at
speeds of approximately 10 and 100 .feet per second. Emphasiswas placed on turbulence and conditions near the wall. Theresults include relevant mean and statistical quantities, such as
Reynolds stresses, triple correlatzons, turbulent d_ssipation, andenergy spectra. It is shown that rates of turbulent-energy
production, dissipation, and diffusion have sharp maximums
near the edge of the laminar sublayer and that there exist a
strong movement of kinetic energy away from this point and anequally strong movement of pressure energy toward it. Finally
it is suggested that, from the standpoint of turbulent structure,
the field may be divided into three regions: (1) Wall proximitywhere turbulence production, diffusion, and viscous action are
all of about equal importance; (2) the central region of the pipewhere energy diffusion plays the predominant role; a_d (3) the
region between (1) and (2) where the local rate of change ofturbulent-energy production dominates the energy received by
diffusive action.
INTRODUCTION
The one aspect of turbulent shear flow that stands out
most prominently is thetransport of stream properties byturbulent motions. The transfer process is fundamental, for
it not only shapes the mean-flow field through momentumtransfer but supplies the mechanism by which turbulent
motions receive energy from the mean-flow field. The well-
known phenomenological theories were the first attempts to
give analytical forms for the transfer mechanism by somesimple physical considerations and thereby succeeded in
making predictions about the nature of the mean-velocityfield. Subsequent experiments, however, have clearly dem-
onstrated the inadequacies of these theories, and in a sys-tematic discussion Batchelor (ref. 1) has pointed out theinconsistencies and unreal consequences of the assumptions
involved. Recently, Rotta (ref. 2) and Tchen (ref. 3) pre-
sented more extensive and deeper analytical treatments of
the nonisotropic turbulence problem, and, although their
results show some agreement with the existing experimental
findings, they cannot escape some arbitrariness in assumingthe nature of the transfer mechanism. It is the general
opinion among investigators in this field that more extensive
experimental work on the turbulence mechanism is necessary
1 Supersede,_ NACA TN 2954, "The Structure of Turbulence in Fully Developed Pipe Flow" by John Laufer, 1953
296687--55--2
before a satisfactory analytical formulation of the problem
will be possible.In the last few years various experimental investigations
were carried out in different types of shear flows. It became
apparent that the dynamic and kinematic processes govern-
ing such flows may be quite different. The difference be-tween the so-called free turbulent flows and flows past solid
boundaries has recently been emphasized by the phenomenonof intermittency, first noted iu a jet by Corrsin (ref. 4) and
later more clearly recognized and studied in detail in the
wake of a cylinder by Townsend (ref. 5). It is apparent bynow that in flows like wakes, jets, and those near the free
surface of the boundary layers the intermittency is present
and seems to play a very important role in the transfer
mechanism, while in pipe and channel:flows it is completelyabsent.
Fairly large amounts of experimental information have
been gathered about flows with free boundaries (refs. 6, 7,
and 8); especially extensive is the work by Townsend in aturbulent wake (ref. 5). The recent Work in two-dimen-sional channel flow by the present writer (ref. 9) brought out
some significant features of flows with solid boundaries, bu_the information was far from being complete. One conse-
quence of this study was the realization that in order toobtain a complete picture, say of the turbulent-energy
balance, a knowledge of flow conditions in close proximity
to the wall was of utmost importance.Since no similar investigation, with emphasis on the _urbu-
lent structure, has ever been carried out in fully developed
pipe flow, the present investigation was undertaken. A pipealso offered the simplicity of axially symmetric mean flowand a turbulent field nonhomogeneous in one direction only.
At the same time it afforded an experimentally convenient
setup for obtaining a shear flow of large scale. .4. diameterof 10 inches was chosen as sufficiently large for the favorable
application of hot-wires and for bringing the region near thewall within practical reach. To cover the desired range,
two working Reynolds numbers were chosen, these being
500,000, based on the diameter and an airspeed at the centerof 100 feet per second, and 50,000 at a speed of 10 feet persecond. The higher Reynolds number minimized viscouseffects while the lower one magnified the extent of the
predominantly viscous layer near the wall. This turned outto be a fortunate choice since experimental difficulties and
4 :/I ....... ....
2 REPORT l174--=NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS
errors encountered in one case were usually absent in theother.
The purpose was to investigate the nature of turbulence
and its relation to the mean flow, in particular the rates of
transfer, diffusion, and dissipation of energy. To this endmeasurements were made of the relevant mean and statis-
tical quantities, including their spatial distributions from
within the laminar sublayer to the center and energy spectraat several radii. Thus the purpose was to reveal details offlow structure that could not be found from studies of mean
flow alone, and yet stop short of a direct attempt to find thenature of the transfer mechanism. It is hoped, however,
that advances toward the latter goal will have been made_hrough the present work.
The present investigation was conducted under the spon-sorship and with the financial assistance of the National
Advisory Committee for Aeronautics. The author wishes
to express his gratitude to Dr. G. B. Schubauer for his
constant interest and encouragement throughout this work.The valuable discussions with Dr. C. M. Tehen and Mr.
P. S. Klebanoff are also much appreciated. Thanks are due
to Miss Z. W. Diehl and Mr. K. D. Tidstrom who carried
out the numerical computations and preparation of thereport.
SYMBOLS
a ' pipe radius, 4.86 in.
Do, Dr dimensionless turbulent-kinetic-energy dif-fusion rates;
a 1 d (u'_+_-f-w_) .D°-_---U-_3-r dr rv - ,
el, e2
l_u(kl)
Fo(kl)
(GD),
k.
P
(PD)o, (PD),
Dr-- _4 r drrV - - .
voltage fluctuations across hot-wires
fraction of turbulent energy _'7 associated
with ki, cm3/sec _
fraction of turbulent energy _ associatedwith kl, cm3/sec _
fraction of turbulent energy _ associatedwith k_, cma/sec 2
dimensionless gradient diffusion rate of
turbulent kinetic energy,
r _ 1 d d u2+v2-4-w 2U_4 r d--rr dr 2
one-dimensional wave number in direction
of mean flow, cm -_
mean pressure at any point i'n pipemean pressure at exit of pipe
dimensionless turbnlent-pressure-energy dif-fusion rates;
(PD)o_ _3 1 d r_p, (PD).=-- _ 1 d r_pr dr _4 r dr
(Pr)o, (Pr)_
P
qR
Rr
r'--a--rr*
t
U
uo
u_
Ul, Us, U3
U, V, W
U t, V t , _1)p
w
W_,W_
X
P
lr
P
dimensionless turbulent-energy production
u--_ dU ,_, _-_ dUrates; (Pr)o-a _-_, _rr)_ _
instantaneous value of pressure fluctuations
dynamic pressure at pipe center
Reynolds number based on diameter of pipeand velocity at pipe center
correlation coefficient of u-fluctuations at
two points displaced in a radial direction
coordinate in radial direction; r=0 corre-sponds to pipe center
friction distance parameter, r' U_/_time
mean velocity at any point in pipe
maximum value of mean velocitydU
friction velocity; U_2- -- v (--_-)_ffi_
total (mean-plus-fluctuating) velocities in x,
r, and _ directions, respectively
instantaneous values of velocity fluctua-
tions in x, r, and _ directions, respectivelyroot-mean-square values of velocity fluctua-
tions in x, r, and _ directmns, respectivelymean velocities in radial and azimuthal
directions, respectively
dimensionless turbulent-energy dissipationrates;
Wo =- a _ \bxff \i_xjJ
where u_ refers to three velocity fluctua-tion components u, v, and w and x_, tothree coordinates; repeated indices indi-cate summation
dimensionless direct-viscous-dissipation rate,
U_' \ dr J
dissipation length parameterkinematic viscosity of air
total (mean-plus-fluctuating) pressureair densityazimuthal coordinate
ANALYTICAL CONSIDERATIONS
EQUATIONS OF MOTION OF PIPE FLOW
The continuity equation and Reynolds equation in cylin-drical coordinates for incompressible mean flow have thefollowing forms:
bU _.l brV. 1 bW__ 0
rg_u+v_u+w_u= 1 _Pbx br r b,p p bx
b r_-__4- lb _'_)+_V2U
U -_-_-I- V-_+ r b_ r = p br
1 b r-__t_l b "_.___7_)+_ {v_v v 2 _w_
aW.yaW. Waw+ yw laP /_,u -_-_- --_-_-T --_-_ r - --_ _-_ _ _ ±
THE STRUCTURE OF TURBULENCE IN FULLY DEVELOPED PIPE FLOW
The boundary conditions are at r=0
_'_=0
whereV2 b 2 , 52 . 1 b 1 b 2
=5-_-_-r _-_ r__
(2)
In a fully developed turbulent pipe flow the conditions are
(a) V=O and W=O
_=0(b) b_
(c) The velocity field is independent of the coordinate x
Equations (2) therefore become
1 .bP_ 1 d r_-_-I- v ('d2U+ 1- dU)p bx r "-_ \ dr 2 r
l bP l d -- "_- rv2+ rp br r--_
d-- 2-_'_0=-g; vw-- V-
Integrating the last equation and using the boundary condi-tion _-_=0 at r=a it follows that _"_----0 for all values of r.
.Thus the Reynolds equations for the turbulent pipe flow
reduce to
l bP l d (_-_ dU) (3a)p bx=--r-f_ r --_ --_
1 bP 1 d -- w __. p br=--r-_ rv_-t---r - (35)
Differentiating equation (3b) with respect to x one finds
b_P/brbx=O. Thus bP/bx is independent of r and equa-
tions (3a) and (3b) readily integrate to
g -p b--x=--r --_ -_ +A(x) (4a)
P-- -_-[- _ r w2--'v_ dr+ B(x) (4b)-p j _ r
3
and at _=a
dU--=0dr
u"-g= 0
V2=0
Also let P=Oat x=O and r=a.
A(z) =0,
1 bP 2 U, _p bx a
and, integrating,
Then from equation (4a),
P 2- U,_x+C(r) (5)
p (_
2 U_x" The equationsFrom equations (4b) and (5), B(x) =--a
finally become
dU+r_-__=u-_- _ U__ (6a)
. Frv"_--w-'_ P 2-t-Ja r dr ..... p a U'2x (6b)
It is interesting to note, as already pointed out by Kamp6de F6riet in the case of flows between parallel walls (ref. 10),
that the shearing stress _-_ and the mean velocity occur
together in one equation and the normal stresses v _ and w _and the mean pressure occur together in the other. From
the experimental point of view this has the advantage that
the equations furnish a method of checking the absolute
accuracy of the measurements of the Reynolds stresses if themean-velocity and pressure distributions are known.
ENERGY EQUATIONS
The momentum equations may be written in the form
bu'2bz_---_-_bu'u2rl bu_q_3b_{ u_u2____=_pl_x2r_V2u _
bu_u2 . bu_2__ 1 bU2U3 LU22--U32 1 b_r__ _-_-r-- r _ __=-_-
bUlU 3 , 5U2_._3__ 1 5U32 ]_2'/_2q_3 " 1 b_r{_bx P_tr-_--_= pr _)_
/__ u__ 2 bu3\\v _-7_t_ -_-_)
(7)
4 REPORT l174--NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS
Multiplying the three equations by ul, u2, and ua, respec-
tively, one.'gets
bul 3 bul2u2 1 _UI2U3 _12U2 2 C)TI"
bUlU22 _U23 1 bU22U34 U2(U22--2U32) 2 blr ,
bX _---_--'_ r b_ ' r =--pU2_-Tt
, rl__ 2 ['bu2\ 2 fbU2 "_2 1/'()U2 \2 U2 2 2 bU3-I
_V_tlU32 __ _IU2U32 _. 1 bU33.__ 3U2U32 2 U 3 5W
2 r1__ _ /bu_v [bmv I/bu_\ 2 u32__2 bu3-]
Introducing the velocity and pressure perturbations
u_= UNu
q-t2-----V
_3=W
_=P÷p
averaging, and adding the three equations there is obtainedthe energy equation for the turbulent motion
_dU. 1 d (u_--F_-Fw 2uv-3v÷ r - _3_rv _-P)=
\bxj} (8)
where in the last term u_ refers to the three velocity fluctua-tion components and x_, to the three coordinates; the repeatedindices indicate summation.
The second term on the right-hand side of the equationmay be neglected compared with the other terms for all
values of r, since (a) near the center it approaches zero as
r-->0 as may be shown from the continuity equation; and (b)everywhere else its order of magnitude _"_/D _, where D is a
characteristic length of the order of the pipe radius, issmaller than, for instance, that of the last term of the
equation; that is, _"_/X_>>_"u=/D _, since X2/D_<<I.The equation may now be rewritten as
rv ( -us+ v_+w_+ P'_
rdrrdr 2 _ _ =0 (9)
The energy equation thus obtained has essentially thesame form as that given by Van Kdrm_in for the case of a
turbulent flow between two parallel plates (ref. 11). Thefirst term corresponds to the rate of production of turbulent
energy by the action of the shearing stresses. The second
term represents the rate of energy change due to transfer of
both kinetic and potential energies by the radial velocityfluctuations and is usually referred to as the diffusion term.
The third expression may be regarded as a gradient type of
energy diffusion and is important only very near the wall.The fourth term expresses the rate of energy dissipation into
heat by action of the viscosity.
EQUIPMENT
WIND TUNNEL
The investigation was carried out in the experimental
setup shown in figure 1. A centrifugal blower having a
capacity of 12,000 cubic feet per minute and powered by a15-horsepower constant-speed motor was placed inside a
large pressure box. The box was octagonal in cross sectionand was l0 feet high and 16 feet long. The air was dis-
charged from the blower into the space surrounding the
blower and then passed through a honeycomb, three screens,and a large contraction cone. The honeycomb consisted of
hexagonal cells 2 inches across and 8 inches long. The"screens had 24 meshes per inch and a wire diameter of 0.0075
inch. The contraction cone, which was made out of hard-
wood and was circular in cross section, reduced the air pas-sage from 48 inches to 18 inches ia diameter. The cross
le radialinletvanes
..... Box octanonal in cross section
. - - _........ Centrifugal blowerI
i-.-J
Ill_TM
....... _S_c!e_e___s.......
,18" OD. steel pipe 3' long
,,'" ..?Turning vanes
_" i" .- Honeycomb
,,,'"//,-Roughness ;>-_ long
======================: : : : ;i 'Screen , ,, ""Pressure tops
'Flexible pipe _\ "'10" O.O. brass tubing 16' long"10" O.D. steel pipe 25' long
FIGURE 1.--Schematic diagram of test setup. -
n
o I
i
?i.
THE STRUCTURE OF TURBULENCE
section was then further reduced from 18 inches to 10 inches
in diameter through a 90 ° elbow. After the elbow a short
elastic coupling was used to prevent the transmission of-.vibration from the pressure box to the pipe.. In order to
minimize any flow irregularities due to the elbow, anotherscreen and an 8-inch-long honeycomb were installed in the
entrance sections of a 25-foot-long steel pipe. Furthermore,in order to accelerate the boundary-layer growth, the pipe
wall was artificially roughened by gluing floor-sanding paperto the surface along a length of 2_ feet. With this arrange-ment it was found from the measured mean-velocity distribu-
tion at the end of the steel pipe, that is, after an entrance
length of about 30 diameters, that the flow was fully devel-
oped turbulent. Following the steel pipe, a 16-foot-longseamless brass tube was attached having an inside diameter
of 9.72 inches. This served as the actual test section.
The speed of the tunnel was regulated by throttling theintake of the blower with adjustable vanes and by venting
the pressure box. In the early stages of the investigationit was realized that for the very low velocity pipe experi-
ments, where the intake of the blower had to be nearly
closed, the heat dissipated in the pressure box was sufficientto affect the flow in the test section. This difficulty was
overcome by supplying the air to the system by a ventilating
fan placed on the side of the box as shown in figure 1.
HOT-WIRE EQUIPMENT
The basic hot-wire equipment used during the investiga-tion is described in detail in reference 12. The hot-wires
were made of platinum drawn by the Wollaston process.
Later during the investigation platinum-rhodium wires (90
percent platinum and 10 percent rhodium) were used andwere found to be very satisfactory. The wire diameters were
generally 0.0001 inch; only when the noise problem wasvery critical were finer wires (0.00005 inch) used. For themeasurements of the longitudinal components of the velocity
fluctuations the length of the wires ranged from 0.01 to0.025 inch.
Special care was taken in building the X-type of wires forthe measurements of the cross components v and w. In
order to minimize wire-length effects and to l_e able to work
very near the wall, the size of the wire holder had to be assmall as possible. With the use of prongs made of fine
jeweler's broaches the dimensions of the holder head werecut flown to approximately 0.015 inch by 01.005 inch, the
wires having a length about 0.025 inch.
TRAVERSING MECHANISM
The traversing mechanism simply consisted of a microm-
eter screw on which the hot-wire support was fastened.
The support could be rotated in a plane perpendicular tothe air flow so that the hot-wire could be adjusted parallel
to the wall.The zero reading of the traversing mechanism (r':0)
was found by placing the hot-wire close to the wall (approxi-
IN FULLY DEVELOPED PIPE FLOW" 5
mately 0.01 inch away) and measuring the distance betweenthe wire and its image in the polished wall by an ocularmicrometer. Since the curvature of the wall was small, the
space between the wire and the wall could be taken as halfthe observed distance.
PROCEDURE AND RESULTS
While fully developed turbulent flow was readily obtained
throughout the 16-foot test length, tedious and time-.consuming adjustments had to be made to remove secondary
effects and obtain axial symmetry. The task was particularlydifficult at the low Reynolds numbers where the effect of
temperature gradient was most felt. This as well as dis-turbances from the elbow had to be eliminated. After the
final adjustments, pressure and velocity surveys made alongthe 16-foot test length showed that the velocity field at each
section was the same and symmetry about the axis had beenobtained. All the final measurements were made inside the
tube 2 to 4 inches from the exit.
MEASUREMENT OF MEAN VELOCITY AND PRESSURE
The experiment was conducted at two Reynolds numbers,
50,000 and 500,000, corresponding to maximum mean veloci-ties of approximately 10 and 100 feet per second, these
velocities being varied slightly in order to compensate for
the daily changes in air viscosity and density. In order tocover the wide range of static and dynamic pressures, aninclined manometer 5 feet in length was used. With benzol
and a slope of 10 to 1 the manometer was sensitive to 0.005centimeter of water by direct reading. With a traveling
microscope and separate scale the sensitivity was increasedto 0.0001 centimeter of water.
The pressure distribution in the direction of the flow was
measured through pressure taps located every 2 feet alongthe brass tube. The results are given in figure 2.
.28
.24
.20
.16
.12
.08
.04
0
// J
,,o/
J
R
o 500,000o 5QO00
I.14 166 8 I0 t2 18
x/2o
FIGURE 2.--Mean-pressure distribution along pipe axis.
6 REPORT l174--NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS
Small total-head and static tubes were made of 0.04-inch-diameter nickel tube stock with 0.003-inch wall thickness.
The tip of the total-head,tube was flattened to an opening of0.006 inch. The static tube was placed approximately 0.4
inch above the total-head tube. This arrangement was used
for measuring the complete velocity distribution in the lowReynolds number case. In the vicinity of the wall a cor-
rection had to be made because of the large level of _elocity
• fluctuations. The correction, having the form
Uoo- =U/l_/ u_+v_+w2 _. /-_-_
-was of the order of 5 percent or less.
1.0
.E
R
u/u,, ,, 5oo,ooo• o 50,000
0 .2 .4 .6 .8 1.0r7o
FIGURE &--Mean-velocity distribution.
.7
r....._ __..__. _o
f / I:_ f
u_u_.3 /R
.2 o 500,000._
o 50,000.I
0 .004 .008 .012 .016 .020 .024 .028 .052 .056 .040 .044r'/O
FIGURE 4.--Mean-velocity distribution near wall Dashed lines are
computed from pressure-drop measurements.
For the high Reynolds number a hot-wire was used to
explore the mean-velocity distribution near the wall. Here
again a correction for velocity fluctuations had to be madebecause of the nonlinear behavior of the hot-wire. This was
accomplished by an approximate graphical method using
the known static response curve of the wire (voltage against
velocity) and the known root-mean-square value of thevoltage fluctuations. The maximum correction was about
10 percent, the correct mean velocity being higher than the
observed. At the low Reynolds number the velocity profiledetermined by the hot-wire was unreasonably low in the
vicinity of the wall after applying the correction. Except
for this, the agreement between the hot-wire results andthose obtained with the total-head tube was good.
The measurements are presented in figures 3 and 4. In
figure 4 the dashed lines indicate the wall velocity gradients
computed from the pressure drop. The agreement with themeasured values is satisfactory.
2.8
2.4
2.0
1.6
u?U r
1.2
.8
A
0 .I .2 .5 .4I
.5 .6r?o
R
a 500,000
o 5i,000
FIGURE 5.--_U p distribution.
.7 .8 .9 1.0
2.8
2.4
2.0
1.6
u/U,
I._)
.8
.4
0 .01
-.e..O
.O2
iD5 .04 .05 .06
rTa.O7
FIGURE 6.--U _ distribution near wall.
Ro 500,000 __o 50,000
.08 .09 .10
THE STRUCTURE OF TURBULENCE IN FULLY DEVELOPED PIPE FLOW
1.2
.8
v_.4
0
1.6
1.2
w_'u,.e
.4
0
R[] 500,000. --o 50,000
.I .2 •:3 .4- •5 •6' .7 °8 .9 I•0
rio
FIGURE 7.--V p and w' distributions.
1.0
.6
b-VIU_.4
.2
O o
.4 _
6-Vlu'v '.2
0 •l
0
IR
[] 500,000o 50,000
"--4
.2 .3 .4 .5 .6 .7 .8 ,9 1.0r2e
FIGUR_ &--Reynolds shearing stress and double-correlation-coefficientdistributions. Curves calculated from measured dU/dr, u', and v'.
MEASUREMENTS OF TURBULENCE LEVELS AND SHEARING STRESS
The three components of the velocity fluctuations _', vp,and w' and the turbulent shearing stress _ were obtained bystandard techniques described in reference 6.
Figures 5 to 8 give the detailed results. In figure 8 thesolid lines represent distributions calculated by equation(6_) using the independently measured mean-velocityg:adients and u' and v'. It is seen that for the high Reynoldsnumber the agreement between these and the directlymeasured points is very good while for the low Reynoldsnumber the measured points are somewhat higher. Itshould be mentioned that lower accuracy of all the measure-ments in the low Reynolds number flow is to be expectedmainly because of the difficulty of forming consistent timeaverages because of the inherent low-frequency fluctuations.
A ==:=_
O o
R500,000
[]
d d.
,u ,o
I I • I.2 .5 .4 .5 .6 .7
rye
5o0o0 _v50 ,_f_V'
U2V
vf_u_v,
w2v_--_v'
I.8 .9 1.0
FIGURE 9.--Triple-correlation coefficients.
.4"
i
o
o
N,> o
i_ 0 [] 130
0
[] E
[] E
0 C
II
.I .2 .5 .4 .5 .6 .7 .8rYa
FIGURE 10.--Quadruple-correlation coefficients•
R w
" 500,000o 50,000
.9 1.0
MEASUREMENTS OF TRIPLE AND QUADRUPLE VELOCITY CORRELATIONS
_leasurements of tbe triple and quadruple velocity corre-lations are given in figures 9, ]0, and 11. The basic techniquedescribed in reference 12 was essentially adopted for thesemeasurements• The schematic diagram of the electroniccircuit is given below:
e2-1Amplifier 2 _ . "_Squaring circuit
8REPORT 1174--NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS
The two hot-wire signals el and e2 are fed into two identical
compensated amplifiers after which e_ is squared• The pre-liminary squaring circuit uses the nonlinear characteristics
of two balanced triodes as employed in Townsend's circuit
(ref. 13). The output of the mixing circuit, which is a
simple resistance network (ref. 12), gives simultaneously thesum and difference of the inputs. Squaring circuits 1 and 2
contain a series of diodes properly biased to give the square
of the inputs. If the signals before the mixing circuit areequalized the ratio meter reads directly the correlation
coefficient _/_/_
. It is seen that in order to obtain the triple velocity correla-
tions by this method the mean fourth power is necessary.
This could have been avoided, of course, if instead of usingthe ratio meter the outputs of the two squaring circuits had
been recorded separately and then subtracted. It w_s felt,however, that the method adopted gave more consistent
results. The mean fourth power was obtained simply byfeeding the amplified hot-wire signal into a squaring circuit
and then squaring the output again by reading the finaloutput on a thermocouple meter.
Unfortunately, for the lower Reynolds number some in-
consistency in the measurements was found using the abovetechnique. In comparing the values of the double-
correlation coefficient _'Y/u'v' obtained by this setup (without
the preliminary squaring circuit) with those obtained by the
conventional method, they were found to be from 20 per-cent to 30 percent lower everywhere except near the wall.
At the high Reynolds number the agreement was very
satisfactory• Although, because of time limitations, it was
LL5
_. _ 500,000 __
.__--_-_/_3 5o,ooo
-..,......_
-I.0
0
-I.5
_/u 3
-J.o _'_ /u,3
\
"50 I .2 .3 .4 .5 .6 .7 .8 * .9rTo
Flcm_. 11._Triple-vetocity-co_elation d_trJbutions.
====._T_
1.0
not possible to trace definitely the cause of this discrep-ancy, it is believed to be due to a difference in the low-
frequency phase shift of the two amplifiers• Consequently,
the large difference between the triple-correlation dis-
tributions at the two Reynolds numbers in figure 11 shouldbe considered to be due to experimental error.
MEASUREMENTS OF VARIOUS DISSIPATION TERMS
The expression for the rate of energy dissipation as itappears in the turbulent-energy equation (9) has the fprm
_ _ _u _ _v _ bw 2+1 /bu", _ i by 2. 1 /bw\2"l
The first three terms were obtained by the differentiation
method introduced by Townsend (ref. 13). By making theassumption
b .I bb-5=-_ a_ (11)
he writes
Thus by electronically differeDtiating the hot-wire signal
(or ,, or\bx) may be easily obtained• A recent
paper by Lin (ref. 14) shows that this assumption is validif no mean-velocity gradient exists and (u/U)2<<l. In ashear flow he gives an additional condition
bu bUU _-_ >>v s_
In the present measurements these conditions were found to
be satisfied with the exception of a region inside the laminar
sublayer. Also during the course of a bound:_r._-layerinvestigation at the NBS an experimental verification of the
validity of this method was made at 0.05 times the boundary-layer thickness.
The fourth and seventh terms of equation (10) were
obtained essentially by a method first suggested by Taylor(ref. 15), in which
_uu(F) 1 [- ,_" "2u 2
as _--+0, where u and u(r-) denote fluctuations at distance F
apart. Thus by measuring the correlation coefficient R, for
small values of F, \br} can be calculated. During the
present investigation the accuracy of the method was
greatly im0roved by adopting the following technique:
THE STRUCTURE OF TURBULENCE IN FULLY DEVELOPED PIPE FLOW9
The above equation may be rewritten as
R u_-}-u_ [u--u_]_- 1 1__ (5u'_"--'--t_
Since, for small values of F, u2=u_(r-), one has
[u-u®] _ 1 7--_-_" r
Several values of a can be easily measured by placing two
hot-wires at various small distances F apart. Then \br]
may be calculated from the slope of a straight line in a plot
of ¢ against _.The remaining four terms of equation (10) could also have
been obtained with the method described above. However,
the wire arrangement necessary for such a technique made
its application impractical. Comparing the first three termsit was found that they satisfy the isotropie relations fairly
well except near the wall (figs. 12 and 13). The other twomeasured terms, while not too different from the first threein
(the center region, are considerably higher especially \br ] ]
near the wall. This, of course, is not surprising since dis-
sipation lengths or microscales in the radial direction are
expected to be smaller because of the presence of the wall.For want of a better procedure it was assumed that mean-
square derivatives with respect to a given direction sepa-
rately satisfied the isotropic relations
1.4xlO 4
1.2
I.C
.e
.E
.4
.2
0
r_\_/=P \bT_) =r _\_/
Moximum ot 5.5 04
-- o \ax/
,__ 2rZu'Z
\ i l
., .o ,.o/In
FIGURE 12.--Distributions of dissipation terms, R= 50,000.
41
.5
.I
0
___AXl041 v\
I I I i I
,- 2U_- " 2UT_
o_ (____'_Z, ,-- v 2r2Ur2 \ 8¢p]
I
.I .2 .5 ,4 .5 .6 .7 .8 .9 1.0r'/o
FIGURE 13.--Distributions of dissipation terms. R=500,000.
It is not possible to estimate the accuracy of this assump-
tion at particular points. However, one can determine theerror made in the total energy dissipation of the pipe cross
section. Integrating the energy equation (9) across the pipe
there is obtained
2_r _a __ dU dr 2_r _a Wr dr (12)
which states that the total energy produced at a given sec-
tion is dissipated at the same section because of the homo-
geneity of the field in the x-direction. Equation (12) wasfound to be satisfied within 10 percent. It is remarked that
if the isotropic relation between all terms had been assumed
and the dissipation calculated in the usual way using
W=15_kbx]
the right-hand side of equation (12) would have been smaller
by approximately a factor of 2.5. It is still an open ques-tion, howe_er, how accurate the approximation is from point
to point.
ENERGY-SPECTRU_M[ _I[rASURE1M[ENTS •
The amplified hot-wire signal was fed into a Hewlett- "Packard wave analyzer with a frequency range of 10 to 16,000
cycles per second. The analyzer selectivity characteristicswere obtained by calibration; two fixed band widths were
chosen, their effective values being 11 and 42 cycles per
second approximately. The output of the analyzer was fedinto a thermocouple circuit. In order to make the measure-
ments independent of the amplifier frequency response the
thermocouple output readings of the hot-wire signal were
matched by a known sine-wave input.The measurements are presented in figures 14, 15, and 16.
Wire-length corrections were applied only to the u'-spectra,
using the method described in reference 16.
10 REPORT l174_NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS
io5
,d
102
I0
<>\-%
\
\
\\
o I.Oo .6910 .28•_ .074v .008:
\
\
\
.o
IOlo-Z iookl, cm-L
FIGURE 14.--ut-spectra. R=500,000.
L01
THE STRUCTURE OF TURBULENCE IN FULLY DEVELOPED PIPE FLOW 11
FIGURE 15.--v'-spectra. R=500,000.
102
REPORT l174--NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS
FIGURE 16.--wt-spectra. R = 500,000.
10z
THE STRUCTURE OF TURBULENCE IN FULLY DE_rELOPED PIPE FLOW"
DISCUSSION
As already pointed out in the "Introduction," the presentunderstanding of the turbulence phenomenon is not com-
plete enough to be able to attack directly the problem con-cerning the nature of the turbulence transfer mechanism
that establishes a stable, nondecaying turbulence field
as the pipe flow. The principal aim of the present work
was an attempt to obtain an over-all picture of the turbulencestructure without trying to understand the detailed mecha-
nism responsible for this structure. To this end the energy_.equations of both the mean and turbulent flow serve as a
useful guide. The former in effect expresses the relative
magnitude of the mean-flow energy loss to the turbulent
field as compared with losses due to dissipation by directmolecular action. The latter gives a relation between the
different forms of turbulent-energy rates such as rates of
production, dissipation, and diffusion. Unfortunately it is
not possible to obtain from these equations any explicit in-
formation on how these various energy changes take place.However, it is hoped that once their relative importance indifferent regions of the turbulent field is better understood
one might attack the ultimate problem of the turbulencemechanism with greater confidence.
A different, but already well-established, method of ap-
proach was also tried. Instead of investigating the different
rates of changes of the turbulent energy at a given point of
the field, its spectral distribution in the wave-number spacewas examined. The merits and shortcomings of this methodwill be discussed later.
91r I iw_. (pr), R.8 o v 500,000
o" of' 50,000
.7
.6
w_,,(pr), .5
.4
!%.0 I0 20 30 40 50 _ 60 70 80 90
r'u_ /.
FIGURE 17.--Comparison of rate of turbulent-energy production with
direct-viscous-dissipation rate near wall
13
MEAN-ENERGY BALANCE
Multiplying the integrated momentum equation (6a) bythe mean-velocity gradient, dU/dr, the following equation isobtained:
r U, 2dU __dU (dU_ _a -_-_uV-_r--_\dr]
Thus the energy available because of the pressure dropalong the pipe is partly converted into turbulent energy and
is partly directly dissipated. The two terms on the right-hand side of the equation are compared in a nondimensional
form near the wall in figure 17. There are two points thatshould be noted here: (a) The bulk of the direct viscous dis-
sipation takes place in a very narrow region, r* _15; (b) the
position where the laminar shearing stress is equal to theturbulent shearing stress (viscous dissipation equal" to turbu-
lence production) is found to be approximately at the same
point where the maximum amount of energy is produced
(r*=11.5). This point is usually referred to as the edge ofthe laminar sublayer. It is seen that not only is the bulk of
the energy taken from the mean flow directly dissipated buta considerable portion of the total turbulence production alsotakes place here.
It is quite apparent from this picture that, in order to
obtain the complete picture of the turbulent-energy balance,conditions near and within the laminar sublayer have to be
known. This, of course, is a very difficult task from theexperimental point of view.
TURBULENT-ENERGY BALANCE
From the measurements presented in the previous sections
all the terms of the turbulent-energy equation (eq. (9)) can
be calculated with the exception of the pressure-energydiffusion. Unfortunately, since this term is small every-
where except near the wall, its determination from equation(9) is very inaccurate. Nevertheless, it is possible to obtain
in a qualitative way an over-all picture of the completeturbulent-energy balance at a cross section. Since the
experimental problems and errors are different in the two
Reynolds number flows they will be discussed separately.
Low Reynolds number flow.--In order to be able to studythe flow conditions very close to the wall it is necessary to
produce as thick a viscous layer as possible. This may bedone by carrying out the measurements at very low mean
speeds. Certain compromises, however, have to be made.First it is much more difficult to establish good flow condi-
tions at low speeds, and second it is more difficult to carryout time-averaging processes in the measurements of statis-
tical quantities. On the other hand the absolute magnitudeof the dissipation terms may be better established since
problems such as wire-length effects, amplifier frequencyresponse, and noise are not so critical.
Figures 18 and 19 give the distribution of all the energyterms in the low Reynolds number flow. It should be noted
14 REPORT 1174--NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS
-24
-t6_
-12
Gain
-8
8
Loss
- (PD)o /
J2 /_'.Wo
20.1 .2 .5 .4 .5 .6
rTo
FIGUR_ 18._Turbulent-energy balance.
are estimated.
.7 .8 .9 1.0
R = 50,000. Dashed lines
(581)-.40
(505)-.52•(PO)r
(229)-.24Gain--i _'_
(152)-.16 L
(76) .08 ,o_J"
Loss
(152) .16(224) .24
0 I0 20 50
/ ..o-----
40 50 60 70 80 90 I00
r'u, /_(.0190) (0581) (0572) (.0762) (.0952)
r'/a
FIGURE 19.--Turbulent-energy balance near wall. R=50,000. Num-
bers in parentheses along ordinate correspond to subscript o, as infigure 18. Dashed lines are estimated.
that in forming dimensionless quantities for the coordinates,
the characteristic length a was chosen in figure 18 and_,/U_ was used in figure 19 representing conditions near the
wall. The correspondence between respective coordinatesis indicated parenthetically in figure 19.
The following interesting conclusions may be drawn:
(1) Throughout the whole cross section, with the exception
of the center region, the rate of energy production at a point
is approximately balanced by the rate of energy dissipation.
(2) All the various energy rates reach a sharp maximumnear the edge of the laminar sublayer.
(3) This edge appears to be also the region from which
the turbulent kinetic energy is diffusing both toward the
pipe center and toward the wall and toward which the
pressure energy is transported.
There is some question about the direction of the pressurediffusion in the center region of the pipe. It was mentioned
in the previous section that the measured triple velocity
correlations are believed to be too low in this region. Assum-
ing that thei_ dimensionless value is approximately the sameas the values found in the high Reynolds number flow--an
assumption which holds approximately true for all other
measured dimensionless statistical quantities--kinetic-energy
diffusion is estimated as shown with a dashed line in figure 18.It is believed that the corresponding pressure-diffusion
distribution (dashed line) comes closer to the actual picture
than that obtained from the directly measured points (solidline). This would indicate that the direction of the pressure
diffusion is toward the luminar sublayer and its value nearthe center is close to zero.
High Reynolds number flow.--As already pointed out,accurate measurements of the various dissipation terms in
the case of high Reynolds number flow become much more
difficult, since, because of the extent of the energy spectrum
to high frequencies, amplifier-tube noise and wire-length
Gain_, (p,-)o
I ..,.._ _
"- (Po),
/)f
8 ,'/
Loss /
I /
//
//
:.,..I
.I .2 .5 .4 .5 .6r'/a
FIeUR_, 20.--Turbulent-energy balance.
are estimated.
.7 .8 .9 1.0
R=500,000. Dashed lines
THE STRUCTURE OF TURBULENCE
effects become increasingly more critical. Although all pre-
cautions were made to minimize these effects, the calculated
rate of energy dissipation is believed to be too small. From
the energy-spectrum measurements it was possible to inferthat considerable dissipation takes place at frequencies as
high as 30,000 cycles per second where the response of the
compensated amplifier ceases to be linear. From the infor-mation gathered in the low Reynolds number flow, an
attempt was made to estimate the dissipation. The follow-
_ing assumptions were made: (1) In the vicinity of the wallwhere similarity with respect to Reynolds number was found
(this will be discussed in detail later) the dissipation can be
.obtained directly from the low Reynolds number measure-
ments; (2) the errors in percentage are the same across the
pipe. This was found to be fairly closely true when thedissipation measurements were repeated using various high-
frequency cut-off filters. With these assumptions and with
equation (12) the estimated dissipation was obtained.Figures 20 and 21 show the directly measured and estimatedvalues. The figures also indicate that the picture of the
energy balance is similar to that obtained in the low Reynoldsnumber flow.
-160
-120
Gain - 80
-40
0
40
Loss 80
120
1600
\\
"_. .(Pr)o
"" (PP)o
/
//l
//
>-..._
.01 .02 .05 .04 .05 .(:}6 .07 D8 .09 .10r?a
FIGURE 21.--Turbulent-energy balance near wall. R=500,000.
Dashed lines are estimated.
ENERGY-SPECTRUM CONSIDERATIONS
_" In the study of isotropic turbulence the concept of the
energy spectrum proved to be a useful and a convenient oneboth from an analytical and a physical point of view. By
considering the distribution of the turbulent energy in thewave-number space it is possible to give in many instances a
simple picture of some basic physical processes taking placein the field. An example is the commonly accepted picture
of the energy transfer from smaller to larger wave numbers.The utility of the spectrum function is much more re-
stricted in a nonisotropic turbulence field where it cannot be
contracted into a scalar, as in the isotropic case, but has tobe treated as a tensor quantity. A restrictive factor in all
cases, and particularly in nonisotropic turbulence, is the factthat measurement is possible of only the so-called Taylor, or
IN FULLY DEVELOPED PIPE FLOW 15
time, spectrum. This means that, even if Taylor's hypoth-
esis (eq. (11))is fulfilled, only a spatial energy distributionover a surface #t = Constant can be measured. This, of
course, prohibits detailed knowledge of the point-by-point
energy distribution in the wave-number space. Neverthe-less, it is possible to draw some conclusions from this type of
measurements.The measured u'-spectra (fig. 14) indicate that the spectral
distributions obtained not too close to the wall manifestsimilar behavior over a wide range of wave numbers. Specifi-
cally, they vary as the -5/3 power of the wave number k_over a considerable range. This is clearly indicated in
figure 22 where the function #l_l_F_(k_) is shown to be aconstant over the range 1_k_24. The two sets of points
shown in this figure indicate the magnitude of the length
corrections applied to the directly measured values. This,of course, is the same type of behavior as that of the equilib-
rium range of the energy spectrum in an isotropic field first
predicted by Kolmogoroff (ref. 17). One may infer there-fore that for sufficiently large turbulent Reynolds number
flows (u'X/_200 for all measurements) and for the mean-
velocity gradients not too large there exists a wide range ofwave numbers in which the energy represented by the
u'-component is transferred from smaller to larger wavenumbers without being significantly influenced by the tur-
bulent-energy production mechanism or by viscous dissipa- .tion. This apparently is true, even though local isotropy
does no_ exist in this range as will be seen later. It is further
seen that this range varies with r'/a and at r'/a=O.O082,where the mean-velocity gradient is already large, the
spectrum shows a different distribution. It might bementioned that in this case there is a rather wide wave-
number range where the spectrum varies closely as k_ -_ as
predicted by Tchen (ref. 3).
[]
O []
0 o
o Correctedo Uncorrected
[3
0
0 2 .4 6 8 I0 12 14 16 18 20 22
FIGURE 22._Equilibrium range of u'-speetrum. R=500,000;
r' / a=0.28.
The conclusions one may dra_v from the measurements of
the vr- and w_-spectra are significantly different. In order
to clarify this difference a comparison is made between the
v'-speetrum measured at the center of the pipe and thatcalculated from the measured u'-spectrum using isotropie
relations (fig. 23). It is seen ttmt there is a large energy
16 REPORT
deficiency in the low wave-number range, while the energy
content of the higher wave numbers is much larger. Further-
more, the --5/3-power-law type of distribution is completely
missing. Similar statements may be made for the v'-spectra
,o5
l174--NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS
(and w'-spectra) at points other than the center of the pipe.An explanation of such behavior is, of course, difficult
mainly because the one-dimensional nature of the measure-ments conceals the detailed spatial energy distribution.
Measured-- Calculated
i0 _
Fv(kI)
i0 _
00
0 o
0
\
\
\o\
\
0
\o\o
\ 0
\
o
'd,o-2 Id' iO°
k,,c,'n-'
FIeURE 23.--Comparison of measured and calculated v'-spectrum.
l0 t
R = 500,000; r'/a= 1.0.
102
THESTRUCTURE OF TURBULENCE IN FULLY DEVELOPED PIPE FLOW 17
However, it is hard to conceive that the geometry of thefield alone would be responsible for such a behavior. Figure23 may suggest that wave numbers receive a large portion
of their v' energy not only by means of the usual transfer
mechanism from smaller to larger wave numbers. By con-sidering the energy equations for the turbulent components
separately, the term most likely to influence the v'-spectrum
in the wave-number range considered is the pressure termtransferring energy from one velocity component to the
_v
other. Its form p _- suggests that it is associated with wave
numbers larger than those where the bulk of the turbulent
.. energy is produced and smaller than the ones where viscous
dissipation is important. This term therefore may very well
be responsible for the larger v' energy present in the wave-
number range in question. It is difficult to draw any con-clusion about the role played by the diffusion terms; in
general they are thought of as representing a low-frequency
phenomenon and therefore would influence the v'-spectrum
in the low wave-number range only.It should be emphasized that the above discussion is
merely speculative in nature and a considerable amount of
experimental work especially in connection with the pressureterms is necessary to be able to give a quantitative picture
of the energy balance spectrumwise.
GENERAL CONSIDERATIONS
On the basis of the presented measurements the flow field
in the pipe may be divided into three regions exhibiting
different behavior from the point of view of turbulence
structure.: Wall-proximity range.--Measurements near the wall indi-
cate that the well-known wall-proximity law of Prandtl
(ref. 18) for the mean velocity may be extended to thefluctuating-velocity field also. Thus by using U, and v/U,
'i as the characteristic velocity and length parameters, the
various velocity distributions become independent of the
Reynolds number in the approximate range 0 _ r* _ 30(figs. 24 to 26).
The various energy rates as they appear in the equation
of turbulent energy are found to be of equal relative im-portance; their magnitudes are much larger than those in
other regions of the turbulent field and they therefore playa dominant role in the energy balance over the entire field.
16
! 12
{ ulu, 8
J
1 Ra 500,000
o 50,000
?/
I 0 20 50 40 50 60
r'U_/.
FIGURE 24.--Mean-velocity distribution near wall.
_...-o-----
7O
5--
oI.5
D
0 10 20 90
FIGURE 25.--U', V', and w _ distributions near wa_l. :
/f
_z
R uv
-- 500,000---- 50,000
/ I I , I0 10 20 50 40 50 60 70 80 90 100
rU.
FZGURE 26.--Turbulent shearing stress and double-correlation eoeffi-
cient near wall.
Center portion of pipe.--Since the direct effect of viscosity
on the turbulent field is negligible in the center portion ofthe pipe, it may be expected that the distributions of the
fluctuations expressed relative to the characteristic velocity
U, are independent of the Reynolds number. Figures 5, 7,and 8 show this to be the case. •
This region is_further characterized by the fact that it
receives a large portion of its turbulent energy by diffusiveaction.
Intermediate region.--The intermediate range extends
from r*_100 to r'/a_O.1. By considering the turbulent-
energy balance in this region (fig. 21) it is seen that the rateof energy diffusion is much smaller than that of production
or dissipation. This means that the energy produced here
is locally dissipated. It should be noted that this is one of
the implied assumptions of the mixing-length theories.Another assumption, namely, the existence of statistical
similarity between the velocity components, is, however,difficult to accept in view of the discussion in the section
"Energy-Spectrum Considerations." Thus, although the
gradient type of momentum transfer involved in mixing-length theories has more experimental support in this region
than in other portions of the field, its use is not completelyjustifiable.
!
18
SUMMARY OF RESULTS
Fully developed flow in a large pipe was found to provide
a very useful medium in which tc study the structure of
turbulence in shear flow. Embodied in the term "structure"
are the interactions between turbulent motions and mean
flow, and the various transfers and movements of energy
from point to point and from mean flow through the spectrum
of turbulent motions.
The following are the major results:
1. The importance of a detailed knowledge of conditions
in the close proximity of the wall was demonstrated.
2. Using the similarity parameters U_ and _/U_ the flow
field in this region was shown to be independent of the
Reynolds number.
3. The various turbulent-energy rates, such as produc-
tion, diffusion, and dissipation, were found to reach a sharp
maximum at the edge of the laminar sublayer (r*_12),
their magnitude being of equal relative importance but much
larger than those in otber regions of the pipe cross section.
4. It was found that there exist a strong transfer of kinetic
energy a_ay from the edge of the laminar sublayer and an
equally strong movement of pressure energy toward it.
5. In the center region of the pipe, the characteristic
length and velocity• parameters were shown to be a and U,.
6. tn the region of large mean-velocity gradients but out-
side of the dissipative region (between r* _ 100 and r'/a _ O. 1)
energy diffusion was found to be small compared with the
turbulent-energy production.
7. The spectrum measurements indicated that in r_gions
where the mean-velocity gradients are not too large the
u'-spectra vary as the --5/3 power of the wave number over
a considerable wave-number range.
8. A considerably different distribution of the v'- and w',
spectra was measured and an explanation of such a behavior
was attempted.
NATIONAL BUREAU OF STANDARDS,
WASHINGTON, D. C., October 28, 1952.
REFERENCES
1. Batchelor, G. K.: Note on Free Turbulent Flows, With SpecialReference to the Two-Dimensional Wake. Jour. Acre. Sci.,
vol. 17, no. 7, July 1950: pp. 441-445. --
REPORT l174_NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS
2. Rotta, J.: Statistische Theorie nichthomogener Turbuleuz.1. Mitteilung. Zs. Phys., Bd. 129, 1951, pp. 547-572;
2. Mitteilung. Zs. Phys., Bd. 131, 1951, pp. 51-77.
3. Tchen, C. M.: On the Spectrum of Energy in Turbulent ShearFlow. Res. Paper RP2388, Jour. Res., Nat. Bur. Standards,
vol. 50 no. 1, Jan. 1953, pp. 51-62.
4. Corrsin, Stanley: Investigation of Flow in an Axially SymmetricalHeated Jet of Air. NACA WR W-94, 1943. (Formerly NACA
ACR 3L23.)5. Townsend, A. A.: The Fully Developed Turbulent Wake of a
Circular Cylinder. Australian Jour. Sci. Res., ser. A, vol. 2,
no. 4, Dec. 1949, pp. 451-468.
6. Corrsin, Stanley, and Uberoi, Mahinder S.: Spectra and Diffusionin a Round Turbulent Jet. NACA Rep. 1040, 1951. (Supersedes
NACA TN 2124.)
7. Liepmann, Hans Wolfgang, and Laufer, John: Investigations ofFree Turbulent Mixing. NACA TN 1257, 1947.
8. Townsend, A. A.: The Structure of the Turbulent Boundary
Layer. Prec. Cambridge Phil. Soc, vol. 47, pt. 2, Apr. 1951,
pp. 375--395.9. Laufer, John: Investigation of Turbulent Flow in a Two-Dimen-
sional Channel. NACA Rep. 1053, 1951. (Supersedes NACA
TN 2123.)
10. Kamp6 de F6riet, J.: Sur l'_coulement d'un fluide visqueux
incompressible entre deux plaques parall_les ind4finies. LaH'ouille blanche, vol. 3, no. 6, Nov.-Dec. 1948, pp. 509-51'7.
11. Von K_rm_n, Th.: The Fundamentals of the Statistical Theoryof Turbulence. Jour. Aero. Sci., vol. 4, no. 4, Feb. 1937, pp.
131-138.
12. Kov_sznay, Leslie S. G.: Development of Turbulence-Measuring
Equipment. NACA TN 2839, 1953.
13. Townsend, A. A.: Measurement of Double and Triple CorrelationDerivatives in Isotropic Turbulence. Prec. Cambridge Phil.
Soc., vol. 43, pt. 4, Oct. 1947, pp. 560-570.
14. Lin, C. C.: On Taylor's Hypothesis and the Acceleration Termsin the Navier-Stokes Equations. Quart. Appl. Math., vol. X,
no. 4, Jan. 1953, pp. 295-306.
15. Taylor, G.I.: Statistical Theory of Turbulence. III--Distributionof Dissipation of Energy in a Pipe Over Its Cross-Section.Prec. Roy. Soc. (London), ser. A, vol. 151, no. 873, Sept. 2, 1935,
pp. 455-464.
16. Uberoi, Mahinder S., anti Kov_znay, Leslie S. G.: On Mappiugand Measurement of Random Fields. Quart. Appl. Math.,_
voL X, no. 4, Jan. 1953, pp. 375-393.
17. Kolmogoroff, A.: The Local Structure of Turbulence in Incom-
pressible Viscous Fluid for Very Large Reynolds' Numbers.
Comp. rend., Acad. Sci. URSS, vol. 30, no. 4, Feb. 10,1941,
pp. 301-305.
18. Prandtl, L.: Bericht tiber Untersuchungen zur ausgebildetenTurbulenz. Z.a.M. M, Bd. 5, Heft 2, Apr. 1925, pp. 136-139.
U. S. GOVERNMENT pRINTiNG OFFICE:ISS5